Learning to read one or more foreign languages is essential for social and economic success in this era of globalization. Decades of research have provided much insight into the cognitive and neural mechanisms for reading. Nevertheless, questions remain regarding how a key component of reading, i.e., the visual form- sound association, is instantiated in the brain, and how it is modulated by factors such as culture and prior language experience. Through an innovative design, the proposed study will address five research questions regarding the neural bases of form-sound association by overcoming three major weaknesses in previous research. First, because in natural languages the features of the visual forms (e.g., alphabetic vs. logographic) typically determine the principles of form-sound association (e.g., assembled vs. addressed phonology, respectively), existing studies have had difficulty disentangling the effects of visual forms from those of form- sound associations. Second, although recent studies have shown that native languages can constrain the neural mechanisms for the processing of native and second languages, few studies have used a factorial design to evaluate how the nature of both native and second languages determines the neural bases of second language learning. Third, although individual differences in reading ability are evident, few studies (none from a cross-linguistic perspective) have tried to integrate behavioral and neurofunctional data to predict the efficiency in learning to read and comprehend a new language. We have developed an artificial language (based on Korean Hangul) that can be learnt via different types of form-sound associations. In the proposed study with four experiments, we will train 120 American (60 with experience with Chinese language and 60 without) and 60 Chinese adult subjects (all of whom had experience with English) to learn this artificial language. Half of each sample will be trained via the addressed phonology-learning to read 60 Hangul characters that do not follow grapheme-to-phoneme correspondence (GPC) rules, whereas the other half will be trained to learn the Hangul via the assembled phonology. Subjects will be trained 1hr/day for 9 to 10 days to learn simultaneously the visual form, form-sound associations, and semantics of 60 characters. They will be scanned before and after training while performing passive viewing, passive listening, and visuo-auditory integration tasks. The same MRI scanners (Siemens TRIO with TIM system) will be used in the US and China. Careful calibration will be performed to ensure that the brain imaging data are compatible between the two sites. Results of this study will answer the following questions: (1) With the same visual forms, do addressed and assembled phonology rely on different neural networks? (2) How does participants' native language (English vs. Chinese) affect or constrain the neural networks involved in learning the visual form-sound associations (or phonological access) of a new language? (3) How does participants' second-language experience (Americans with or without experience with Chinese) affect the neural mechanisms of phonological access? (4) What behavioral, neuroanatomical, and neurofunctional factors can predict the ability of people with different native languages to acquire form-sound associations under different learning conditions? In addition to these questions, the results of this study will also be compared to those from another related NSF-funded study (a training condition in which visual form and phonology are trained without semantic training). Based on such comparisons, we will address a fifth research question: What is the effect of semanti training on the neural networks of phonological access?